Job-Hunt Agent — A Forward-Deployed Agent, With Me in Command
A forward-deployed agent that hunts — with me in command.
- •Scans official ATS APIs (Greenhouse / Lever / Ashby) — 2,300+ real roles, legally
- •Transparent matching: profile fit + seniority/comp tier + relocation fit
- •Human-in-command cockpit with an in-app command console
- •Novel architecture: the agent’s brain runs on a subscription, not a paid LLM API
Tap to enlargeA personal AI agent that runs my own job search end-to-end: it scans target companies, ranks roles against my profile, and drafts outreach — but nothing reaches a recruiter without my approval. Built as a live proof of the forward-deployed, human-in-command model I advocate: autonomous where it helps, human-controlled where it matters. Roadmap: rebuilding the scanner and matcher as native n8n workflows — turning the orchestration into a shareable, agentic pipeline.
Clean Data, No Scraping
Thousands of real roles, pulled legally through official APIs
The killer insight: the companies I target all sit on a handful of applicant-tracking systems with open JSON APIs — Greenhouse, Lever, Ashby. The scanner pulls thousands of real, current roles legally and for free, then normalizes them into one database. No scraping, no gray areas.
Ranking That Reflects a Real Life
From thousands of listings to the handful genuinely worth pursuing
A generic “good fit” score is useless. The matcher encodes an actual person’s constraints — role axis, seniority level, compensation floor, and relocation requirements — and tags each role with a transparent, tunable reason. It cuts thousands of listings down to the handful genuinely worth the time.
Human-in-Command, End to End
The agent proposes; the human commits — always
The agent drafts; the human decides. A cockpit shows the funnel, ranked roles, and an approvals inbox, plus a live chat console for commands. Nothing is sent, applied, or archived without a click — the same governance principle I build into enterprise agents, applied to my own search.
The Brain Runs on a Subscription
Fully agentic — with zero LLM API keys and zero per-token cost
The interesting architectural choice: instead of wiring a paid LLM API into the app, the intelligence is a headless assistant process running on a subscription — kept warm for fast responses. The app stays a clean cockpit with zero model keys and zero per-token cost, while still being fully agentic.
Interested in building something similar?
Let's Connect